Comment by goosedragons | Hacker News Reader

Comment by goosedragons | Hacker News Reader

Parent Full thread

What does that really mean at this point? A big part of what made integrated cards suck so much is how they had to share memory with the CPU and with a unified memory SiP design that doesn't seem like a problem. Plus the M1 Ultra has 114 billion transistors whereas the 3090 is 28 billion. Even if the M1 Ultra is only using a 1/4 for the GPU that's still the same number of transistors. Even a 6th is still huge.

If Apple wants people to take their claims seriously and if weak engine support for Metal and Rosetta support is a problem they should pay someone to port a popular graphics intensively game properly. Of course it works in their favor not to because fans can point to it and say we'll it's just not optimized for it! Like we all did back in the PPC days until they admitted how crap those chips were when switching to Intel.

Apple doesn't care about gaming, but they do care about 3D rendering. Blender benchmarks (which is just now getting official Metal support, sponsored by Apple), should be the real deal.

Blender is still 3/4 times slower on ultra compared to rtx 3090 according to people testing it

34s on ultra (64 core gpu) vs 11s on rtx 3090

So 3 times slower but taking ~4 times less power (350W 3090 70W ultra [on 48 core gpu so about 90 on 64 cores])

Sounds about right, it's not magic. AFAIK the Metal support in Blender is not very optimized yet (it's brand new) but even optimized I don't expect it to come close to an RTX 3090. I still find the results amazing, for the amount of power this thing needs if it reaches 2x slower after optimizations, that's massive. Don't forget that the Ultra also includes the CPU and RAM in the power consumption.

Its wall power without cpu encoding with cpu its ~140W.

But i can agree results are good but we need also to remember Samsung 8nm [rtx 3090 node] is 4 years old so 2 generations different from TSMC 5nm 4090 will be probably on 5nm node in few months and presumably will have 3x teraflops (probably 3x tf32 precision but it could be fp32 using 2/1.5x(?) more power).

(also I'm not sure about optimization it was write by apple employees and apple likes to drop open source support and focus on proprietary software)

The RAM packaging doesn't really matter much, Apple runs JEDEC spec memory timings anyway.

But it is impressive that they get high GPU performance with LPDDR. Consoles also use one big SoC and unified memory – but with GDDR memory!

Reply on news.ycombinator.com